Taek Mu Kwon

The Efficacy of Red Flags in Predicting the SEC's Targets: An Artificial Neural Networks Approach

This paper illustrates the application of artificial neural networks (ANN) to test the ability of selected SAS 53 red flags in predicting the targets of the SEC investigations. ANN models classify the membership in target (investigated) versus control (non-investigated) firms with an average accuracy of 81%. These results confirm the value of red flags ,i.e., financial ratios available from trial balance in conjunction with non-financial red flags such as the turnover of CEO,CFO, and auditors do have predictive value.

A class of robust entropic functionals for image restoration

This paper considers the concept of robust estimation in regularized image restoration. Robust functionals are employed for the representation of both the noise and the signal statistics. Such functionals allow the efficient suppression of a wide variety of noise processes and permit the reconstruction of sharper edges than their quadratic counterparts. A new class of robust entropic functionals is introduced, which operates only on the high-frequency content of the signal and reflects sharp deviations in the signal distribution. This class of functionals can also incorporate prior structural information regarding the original image, in a way similar to the maximum information principle. The convergence properties of robust iterative algorithms are studied for continuously and noncontinuously differentiable functionals. The definition of the robust approach is completed by introducing a method for the optimal selection of the regularization parameter. This method utilizes the structure of robust estimators that lack analytic specification. The properties of robust algorithms are demonstrated through restoration examples in different noise environments.

A parallel sorting network without comparators: A neural network approach

The design of a parallel sorting network that does not use any comparators is described. The network consists of a two-dimensional array of neural nodes, representing the sorted result through the position of the on-state neuron in each column. The Kth column of the array finds the Kth largest value (Kth winner) and displays it by turning on the Kth winner node. Thus, the sorted result can be read in order of the on-state nodes from the first to the last column. Since the role of the Kth column of the network is to find the Kth winner, each column of the proposed sorting network is referred to as the Kth WTA network. A simple design technique of the Kth WTA network, which can be readily implemented in hardware, is described.<<ETX>>

A multilayered perceptron approach to prediction of the SEC's investigation targets

In the fields of accounting and auditing, detection of firms engaged in fraudulent financial reporting has become increasingly important, due to the increased frequency of such events and the attendant costs of litigation. The neural-network approach sheds some light on this problem due to the attributes that it requires minimum prior knowledge of the data and achieves a highly nonlinear computational model based on past experience (training). In this study, we employ seven red flags which are composed of four financial red flags and three turnover red flags in order to detect targets of the Securities and Exchange Commissions (SECs) investigation of fraudulent financial reporting. The red flags are computed over 70 firms spread among various industrial sectors, and form the base data that is used for developing the computational prediction model. Multilayered perceptron computation of this data was able to predict the targets of the SEC investigated firms with an average of 88% accuracy in the cross-validation test. On the other hand, the same data computed by the logit program gave an average prediction rate of 47%

Contrast enhancement for backpropagation

This paper analyzes the effect of data-contrast to a backpropagation (BP) network and introduces a data preprocessing algorithm that can improve the efficiency of the standard BP learning. The basic idea is to transform input data to a range that associates the high-slope region of the sigmoid function where a relatively large modification of weights occurs. A simple uniform transformation to such a desired range, however, can lead to a slow and unbalanced learning if the data distribution is heavily skewed. To facilitate data processing on such distribution, the authors propose a modified histogram equalization technique which enhances the sparing between the data points in the heavily concentrated regions of the distribution.

Multiresolution image restoration in the wavelet domain

This paper proposes an image restoration approach in the wavelet domain that directly associates multiresolution with multichannel image processing. We express the formation of the multiresolution image as an operator on the image domain that transforms block-circulant structures into partially-block-circulant structures. We prove that the stationarity assumption in the image domain leads to the suppression of cross-band correlation in the multiresolution domain. Moreover, the space invariance assumption leads to the loss of cross-band interference and interaction. In addition to the rigorous explanation of these effects, our formulation reveals new correlation schemes for the multiresolution signal in the wavelet domain. In essence, the proposed implementation relaxes the stationarity and space-invariance assumptions in the image domain and introduces new operator structures for the implementation of single-channel algorithms that take advantage of the correlation structure in the wavelet domain. We provide a rigorous study of these effects for both the equal-rate subband decomposition and the multiresolution pyramid decomposition. Several image restoration examples on the Wiener-filtering approach show significant improvement achieved by the proposed approach over the conventional discrete Fourier transform (DFT) implementation. >

On the application of robust functionals in regularized image restoration

The authors address aspects of robust estimation in regularized image restoration, with the utilization of nonquadratic objective functions. The structural flexibility of generalized maximum-likelihood functions and M-estimators is exploited to provide accurate representation of a wide class of posterior (noise) distribution functions. The utilization of nonquadratic smoothing functionals for the restoration of sharp edges is addressed. In the context of robust estimation, the authors introduce novel entropic functionals that operate on a high-pass version of the original image and can accurately characterize a wide ensemble of images. The entropic functionals permit large signal deviations and enable the reconstruction of sharp edges. The properties of the robust algorithms are demonstrated through restoration examples in different noise environments.<<ETX>>

Gaussian perceptron: experimental results

A new neural model which has a Gaussian activation function is presented. This model is referred to as the Gaussian perceptron. For the training of single-layered Gaussian perceptrons, the Gaussian perceptron learning algorithm, which is a variant of the conventional perceptron learning algorithm, is presented. The winner-take-all algorithm is proposed as a multilayer training algorithm. A number of examples are presented along with the comparison with backpropagation networks, which demonstrate the performance of Gaussian perceptron networks.<<ETX>>

A degenerated fuzzy-number processing system based on artificial neural networks

In this paper, a degenerated fuzzy-number processing system based on artificial neural networks (ANNs) is introduced. The digital representation of fuzzy numbers is assumed, where the universe of discourse is discretized into n equally divided intervals. It is proposed that fuzzy-number processing be performed in two basic stages. The first stage performs the retrieval of fuzzy data consisting of degenerated fuzzy numbers and the second stage performs the desired fuzzy operations on the retrieved data. The method of incomplete fuzzy-number retrieval is proposed based on an ANN structure that is trained to estimate the missing membership function values.

Common Data Format Archiving of Large-Scale Intelligent Transportation Systems Data for Efficient Storage, Retrieval, and Portability

Intelligent transportation system (ITS) sensor networks, such as road weather information and traffic sensor networks, typically generate enormous amounts of data. As a result, archiving, retrieval, and exchange of ITS sensor data for planning and performance analysis are becoming increasingly difficult. An efficient ITS archiving system that is compact and exchangeable and allows efficient and fast retrieval of large amounts of data is essential. A proposal is made for a system that can meet the present and future archiving needs of large-scale ITS data. This system is referred to as common data format (CDF) and was developed by the National Space Science Data Center for archiving, exchange, and management of large-scale scientific array data. CDF is an open system that is free and portable and includes self-describing data abstraction. Archiving traffic data by using CDF is demonstrated, and its archival and retrieval performance is presented for the Minnesota Department of Transportation–s 30-s traffic data collected from about 4,000 loop detectors around Twin Cities freeways. For comparison of the archiving performance, the same data were archived by using a commercially available relational database, which was evaluated for its archival and retrieval performance. This result is presented, along with reasons that CDF is a good fit for large-scale ITS data archiving, retrieval, and exchange of data.